MATH 475 Final Project

Predicting NFL QB All-Pros

Authors: Brayan Mauricio-Gonzalez and Graham Swain

import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split

from sklearn.model_selection import GridSearchCV
from sklearn.linear_model import LogisticRegression
from sklearn.tree import DecisionTreeClassifier
from sklearn.tree import plot_tree
from sklearn.metrics import ConfusionMatrixDisplay
from sklearn.metrics import recall_score
from sklearn.ensemble import RandomForestClassifier
from sklearn.ensemble import StackingClassifier

df = pd.read_csv('nfl_qb.csv')

Looking at the first few rows of our data set to ensure there is nothing that immediately seems wrong.

df.head()

	year	player	tm	age	pos	g	gs	wins	loses	ties	...	rush_rk	rush_att	rush_yds	rush_td	rush_lng	rush_y/a	rush_y/g	fmb	ap_1st	ap_2nd
0	1970	John Brodie	SFO	35	QB	14	14	10	3	1	...	0	0	0	0	0	0.0	0.0	0	1	0
1	1970	Fran Tarkenton	NYG	30	QB	14	14	9	5	0	...	0	0	0	0	0	0.0	0.0	0	0	0
2	1970	Jim Hart	STL	26	QB	14	14	8	5	1	...	127	18	18	0	4	1.0	1.3	6	0	0
3	1970	Roman Gabriel	RAM	30	QB	14	14	9	4	1	...	106	28	104	1	15	3.7	7.4	6	0	0
4	1970	Daryle Lamonica	OAK	29	QB	14	14	8	4	2	...	0	0	0	0	0	0.0	0.0	0	0	0

5 rows × 42 columns

df = df.loc[df['pass_att'] >= 75]

Checking if any cells are N/A.

df.isna().sum()

year               0
player             0
tm                 0
age                0
pos                0
g                  0
gs                 0
wins               0
loses              0
ties               0
pass_rk            0
cmp                0
pass_att           0
cmp%               0
pass_yds           0
pass_td            0
td%                0
int                0
int%               0
pass_lng           0
pass_y/a           0
ay/a               0
y/c                0
pass_y/g           0
rate               0
sk                 0
yards_lost_sack    0
sk%                0
ny/a               0
any/a              0
4qc                0
gwd                0
rush_rk            0
rush_att           0
rush_yds           0
rush_td            0
rush_lng           0
rush_y/a           0
rush_y/g           0
fmb                0
ap_1st             0
ap_2nd             0
dtype: int64

Making a new column 'was_ap' that will be one if the player received 1st or 2nd team AP honors that season. We are using a bitwise or operator to accomplish this.

df['was_ap'] = (df['ap_1st'] | df['ap_2nd'])

This formats all of the Seaborn plots in the manner that we want.

sns.set(rc={'figure.figsize':(20,7.55)})
sns.set_style('ticks')

This sorts the players who achieved AP honors to the top of the dataframe so they will be displayed on the top layer of graphs.

graph_df = df.sort_values(by = 'was_ap')

fig, axs = plt.subplots(nrows = 5)
fig.set_size_inches(15, 15)
fig.tight_layout(pad = -1.5)
axs[0].set_xticks([])
axs[1].set_xticks([])
sns.scatterplot(data = graph_df, x = 'year', y = 'g', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[0])
sns.scatterplot(data = graph_df, x = 'year', y = 'gs', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[1])
sns.scatterplot(data = graph_df, x = 'year', y = 'wins', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[2])
sns.scatterplot(data = graph_df, x = 'year', y = 'loses', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[3])
sns.scatterplot(data = graph_df, x = 'year', y = 'ties', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[4])

<AxesSubplot:xlabel='year', ylabel='ties'>

fig, axs = plt.subplots(nrows = 6)
fig.set_size_inches(15, 18)
fig.tight_layout(pad = -1.5)
axs[0].set_xticks([])
axs[1].set_xticks([])
sns.scatterplot(data = graph_df, x = 'year', y = 'pass_rk', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[0])
sns.scatterplot(data = graph_df, x = 'year', y = 'pass_att', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[1])
sns.scatterplot(data = graph_df, x = 'year', y = 'cmp', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[2])
sns.scatterplot(data = graph_df, x = 'year', y = 'pass_yds', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[3])
sns.scatterplot(data = graph_df, x = 'year', y = 'pass_td', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[4])
sns.scatterplot(data = graph_df, x = 'year', y = 'int', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[5])

<AxesSubplot:xlabel='year', ylabel='int'>

fig, axs = plt.subplots(nrows = 3)
fig.set_size_inches(15, 9)
fig.tight_layout(pad = -1.5)
axs[0].set_xticks([])
axs[1].set_xticks([])
sns.scatterplot(data = graph_df, x = 'year', y = 'cmp%', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[0])
sns.scatterplot(data = graph_df, x = 'year', y = 'td%', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[1])
sns.scatterplot(data = graph_df, x = 'year', y = 'int%', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[2])

<AxesSubplot:xlabel='year', ylabel='int%'>

fig, axs = plt.subplots(nrows = 4)
fig.set_size_inches(15, 12)
fig.tight_layout(pad = -1.5)
axs[0].set_xticks([])
axs[1].set_xticks([])
sns.scatterplot(data = graph_df, x = 'year', y = 'pass_lng', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[0])
sns.scatterplot(data = graph_df, x = 'year', y = 'pass_y/a', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[1])
sns.scatterplot(data = graph_df, x = 'year', y = 'ay/a', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[2])
sns.scatterplot(data = graph_df, x = 'year', y = 'rate', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[3])

<AxesSubplot:xlabel='year', ylabel='rate'>

fig, axs = plt.subplots(nrows = 3)
fig.set_size_inches(15, 9)
fig.tight_layout(pad = -1.5)
axs[0].set_xticks([])
axs[1].set_xticks([])
sns.scatterplot(data = graph_df, x = 'year', y = 'y/c', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[0])
sns.scatterplot(data = graph_df, x = 'year', y = 'ny/a', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[1])
sns.scatterplot(data = graph_df, x = 'year', y = 'any/a', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[2])

<AxesSubplot:xlabel='year', ylabel='any/a'>

fig, axs = plt.subplots(nrows = 3)
fig.set_size_inches(15, 9)
fig.tight_layout(pad = -1.5)
axs[0].set_xticks([])
axs[1].set_xticks([])
sns.scatterplot(data = graph_df, x = 'year', y = 'sk', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[0])
sns.scatterplot(data = graph_df, x = 'year', y = 'sk%', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[1])
sns.scatterplot(data = graph_df, x = 'year', y = 'yards_lost_sack', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[2])

<AxesSubplot:xlabel='year', ylabel='yards_lost_sack'>

fig, axs = plt.subplots(nrows = 2)
fig.set_size_inches(15, 6)
fig.tight_layout(pad = -1.5)
axs[0].set_xticks([])
axs[1].set_xticks([])
sns.scatterplot(data = graph_df, x = 'year', y = '4qc', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[0])
sns.scatterplot(data = graph_df, x = 'year', y = 'gwd', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[1])

<AxesSubplot:xlabel='year', ylabel='gwd'>

fig, axs = plt.subplots(nrows = 4)
fig.set_size_inches(15, 12)
fig.tight_layout(pad = -1.5)
axs[0].set_xticks([])
axs[1].set_xticks([])
sns.scatterplot(data = graph_df, x = 'year', y = 'rush_rk', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[0])
sns.scatterplot(data = graph_df, x = 'year', y = 'rush_att', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[1])
sns.scatterplot(data = graph_df, x = 'year', y = 'rush_yds', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[2])
sns.scatterplot(data = graph_df, x = 'year', y = 'rush_td', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[3])

<AxesSubplot:xlabel='year', ylabel='rush_td'>

df.drop(columns = {'pos', 'pass_rk', 'rush_rk', 'rush_att', 'rush_yds', 'rush_td', 'rush_lng', 'rush_y/a',  'rush_y/g', 'fmb'}, inplace = True)

def normalize_by_year(df):
    def normalize(group):
        for column in group.columns:
            if ((str(group[column].values[0]).isnumeric() or isinstance(group[column].values[0], float))
                and column not in ['year', 'age', 'ap_1st', 'ap_2nd', 'was_ap']):
                group[column] = (group[column] - min(group[column])) / (max(group[column]) - min(group[column]))

        return group

    groups = df.copy().groupby("year")
    return groups.apply(normalize)

df_normalized = normalize_by_year(df)

graph_df_n = df_normalized.sort_values(by = 'was_ap')

fig, axs = plt.subplots(nrows = 5)
fig.set_size_inches(15, 15)
fig.tight_layout(pad = -1.5)
axs[0].set_xticks([])
axs[1].set_xticks([])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'g', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[0])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'gs', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[1])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'wins', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[2])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'loses', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[3])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'ties', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[4])

<AxesSubplot:xlabel='year', ylabel='ties'>

fig, axs = plt.subplots(nrows = 5)
fig.set_size_inches(15, 15)
fig.tight_layout(pad = -1.5)
axs[0].set_xticks([])
axs[1].set_xticks([])

sns.scatterplot(data = graph_df_n, x = 'year', y = 'pass_att', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[0])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'cmp', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[1])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'pass_yds', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[2])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'pass_td', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[3])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'int', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[4])

<AxesSubplot:xlabel='year', ylabel='int'>

fig, axs = plt.subplots(nrows = 3)
fig.set_size_inches(15, 9)
fig.tight_layout(pad = -1.5)
axs[0].set_xticks([])
axs[1].set_xticks([])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'cmp%', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[0])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'td%', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[1])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'int%', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[2])

<AxesSubplot:xlabel='year', ylabel='int%'>

fig, axs = plt.subplots(nrows = 4)
fig.set_size_inches(15, 12)
fig.tight_layout(pad = -1.5)
axs[0].set_xticks([])
axs[1].set_xticks([])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'pass_lng', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[0])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'pass_y/a', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[1])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'ay/a', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[2])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'rate', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[3])

<AxesSubplot:xlabel='year', ylabel='rate'>

fig, axs = plt.subplots(nrows = 3)
fig.set_size_inches(15, 9)
fig.tight_layout(pad = -1.5)
axs[0].set_xticks([])
axs[1].set_xticks([])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'sk', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[0])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'sk%', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[1])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'yards_lost_sack', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[2])

<AxesSubplot:xlabel='year', ylabel='yards_lost_sack'>

fig, axs = plt.subplots(nrows = 2)
fig.set_size_inches(15, 6)
fig.tight_layout(pad = -1.5)
axs[0].set_xticks([])
axs[1].set_xticks([])
sns.scatterplot(data = graph_df_n, x = 'year', y = '4qc', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[0])
sns.scatterplot(data = graph_df_n, x = 'year', y = 'gwd', hue = 'was_ap', size = 'was_ap', size_order = [1, 0], style = 'was_ap', style_order = [1, 0], legend = False, ax = axs[1])

<AxesSubplot:xlabel='year', ylabel='gwd'>

# df.drop(columns = {'pos', 'pass_rk', 'rush_rk', 'rush_att', 'rush_yds', 'rush_td', 'rush_lng', 'rush_y/a',  'rush_y/g', 'fmb'}, inplace = True)
df_normalized.drop(columns = {'year', 'tm', 'age', 'gs', 'wins', 'loses', 'ties', 'pass_lng', 'pass_att',
                              'int', 'sk', 'sk%', 'yards_lost_sack', '4qc', 'gwd', 'ap_1st', 'ap_2nd',
                              'player'}, 
                   inplace = True)

Decision Tree

y = df_normalized['was_ap']
X = df_normalized.drop(columns = {'was_ap'})

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.3, random_state = 12)

DTC = DecisionTreeClassifier(max_depth = 4)
DTC.fit(X_train, y_train)

DecisionTreeClassifier(max_depth=4)

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

print(f"Our model's score against the testing data: {DTC.score(X_test, y_test)*100:4.2f}%")

Our model's score against the testing data: 95.81%

ConfusionMatrixDisplay.from_estimator(DTC, X_test, y_test)

<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7f1de55406d0>

y_pred = DTC.predict(X_test)
print(f"Our model's accuracy for predicting All-Pro's: {recall_score(y_test, y_pred)*100:4.2f}%")

Our model's accuracy for predicting All-Pro's: 44.12%

plot_tree(DTC, feature_names = X.columns, class_names = ['Was NOT AP', 'Was AP'])

[Text(0.5267857142857143, 0.9, 'pass_td <= 0.965\ngini = 0.08\nsamples = 1614\nvalue = [1547, 67]\nclass = Was NOT AP'),
 Text(0.2857142857142857, 0.7, 'pass_td <= 0.759\ngini = 0.041\nsamples = 1563\nvalue = [1530, 33]\nclass = Was NOT AP'),
 Text(0.14285714285714285, 0.5, 'ny/a <= 0.858\ngini = 0.014\nsamples = 1437\nvalue = [1427, 10]\nclass = Was NOT AP'),
 Text(0.07142857142857142, 0.3, 'any/a <= 0.992\ngini = 0.004\nsamples = 1365\nvalue = [1362, 3]\nclass = Was NOT AP'),
 Text(0.03571428571428571, 0.1, 'gini = 0.003\nsamples = 1364\nvalue = [1362, 2]\nclass = Was NOT AP'),
 Text(0.10714285714285714, 0.1, 'gini = 0.0\nsamples = 1\nvalue = [0, 1]\nclass = Was AP'),
 Text(0.21428571428571427, 0.3, 'cmp <= 0.571\ngini = 0.176\nsamples = 72\nvalue = [65, 7]\nclass = Was NOT AP'),
 Text(0.17857142857142858, 0.1, 'gini = 0.039\nsamples = 50\nvalue = [49, 1]\nclass = Was NOT AP'),
 Text(0.25, 0.1, 'gini = 0.397\nsamples = 22\nvalue = [16, 6]\nclass = Was NOT AP'),
 Text(0.42857142857142855, 0.5, 'ny/a <= 0.934\ngini = 0.298\nsamples = 126\nvalue = [103, 23]\nclass = Was NOT AP'),
 Text(0.35714285714285715, 0.3, 'cmp% <= 0.937\ngini = 0.248\nsamples = 117\nvalue = [100, 17]\nclass = Was NOT AP'),
 Text(0.32142857142857145, 0.1, 'gini = 0.207\nsamples = 111\nvalue = [98, 13]\nclass = Was NOT AP'),
 Text(0.39285714285714285, 0.1, 'gini = 0.444\nsamples = 6\nvalue = [2, 4]\nclass = Was AP'),
 Text(0.5, 0.3, 'int% <= 0.313\ngini = 0.444\nsamples = 9\nvalue = [3, 6]\nclass = Was AP'),
 Text(0.4642857142857143, 0.1, 'gini = 0.245\nsamples = 7\nvalue = [1, 6]\nclass = Was AP'),
 Text(0.5357142857142857, 0.1, 'gini = 0.0\nsamples = 2\nvalue = [2, 0]\nclass = Was NOT AP'),
 Text(0.7678571428571429, 0.7, 'rate <= 0.914\ngini = 0.444\nsamples = 51\nvalue = [17, 34]\nclass = Was AP'),
 Text(0.6785714285714286, 0.5, 'int% <= 0.414\ngini = 0.48\nsamples = 25\nvalue = [15, 10]\nclass = Was NOT AP'),
 Text(0.6428571428571429, 0.3, 'pass_y/a <= 0.536\ngini = 0.494\nsamples = 18\nvalue = [8, 10]\nclass = Was AP'),
 Text(0.6071428571428571, 0.1, 'gini = 0.0\nsamples = 4\nvalue = [4, 0]\nclass = Was NOT AP'),
 Text(0.6785714285714286, 0.1, 'gini = 0.408\nsamples = 14\nvalue = [4, 10]\nclass = Was AP'),
 Text(0.7142857142857143, 0.3, 'gini = 0.0\nsamples = 7\nvalue = [7, 0]\nclass = Was NOT AP'),
 Text(0.8571428571428571, 0.5, 'pass_yds <= 0.78\ngini = 0.142\nsamples = 26\nvalue = [2, 24]\nclass = Was AP'),
 Text(0.7857142857142857, 0.3, 'int% <= 0.137\ngini = 0.5\nsamples = 2\nvalue = [1, 1]\nclass = Was NOT AP'),
 Text(0.75, 0.1, 'gini = 0.0\nsamples = 1\nvalue = [0, 1]\nclass = Was AP'),
 Text(0.8214285714285714, 0.1, 'gini = 0.0\nsamples = 1\nvalue = [1, 0]\nclass = Was NOT AP'),
 Text(0.9285714285714286, 0.3, 'pass_td <= 0.974\ngini = 0.08\nsamples = 24\nvalue = [1, 23]\nclass = Was AP'),
 Text(0.8928571428571429, 0.1, 'gini = 0.5\nsamples = 2\nvalue = [1, 1]\nclass = Was NOT AP'),
 Text(0.9642857142857143, 0.1, 'gini = 0.0\nsamples = 22\nvalue = [0, 22]\nclass = Was AP')]

plot_data = X_test.copy()
plot_data['was_ap'] = y_pred
sns.scatterplot(data = plot_data, x = "pass_td", y = "any/a", hue = "was_ap")

<AxesSubplot:xlabel='pass_td', ylabel='any/a'>

Random Forest Classifier (Grid Search)

RFC = RandomForestClassifier()
grid_params = {
    "n_estimators": range(1, 15, 3),
    "max_depth": range(1, 6)
}

RFC_GSCV = GridSearchCV(RFC, grid_params)

RFC_GSCV.fit(X_train, y_train)

RFC_GSCV.best_params_

{'max_depth': 3, 'n_estimators': 7}

best_RFC = RFC_GSCV.best_estimator_
print(f"Our model's score against the training data: {best_RFC.score(X_train, y_train)*100:4.2f}%")

Our model's score against the training data: 97.65%

print(f"Our model's score against the testing data: {best_RFC.score(X_test, y_test)*100:4.2f}%")

Our model's score against the testing data: 96.10%

ConfusionMatrixDisplay.from_estimator(best_RFC, X_test, y_test)

<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7f1de544c670>

y_pred = best_RFC.predict(X_test)
print(f"Our model's accuracy for predicting All-Pro's: {recall_score(y_test, y_pred)*100:4.2f}%")

Our model's accuracy for predicting All-Pro's: 26.47%

Decision Tree (Grid Search)

DTC = DecisionTreeClassifier()
grid_params = {
    "max_depth": range(1, 6)
}

DTC_GSCV = GridSearchCV(DTC, grid_params)

DTC_GSCV.fit(X_train, y_train)

DTC_GSCV.best_params_

{'max_depth': 1}

best_DTC = DTC_GSCV.best_estimator_
print(f"Our model's score against the training data: {best_DTC.score(X_train, y_train)*100:4.2f}%")

Our model's score against the training data: 96.90%

print(f"Our model's score against the testing data: {best_DTC.score(X_test, y_test)*100:4.2f}%")

Our model's score against the testing data: 94.94%

ConfusionMatrixDisplay.from_estimator(best_DTC, X_test, y_test)

<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7f1de536c100>

y_pred = best_DTC.predict(X_test)
print(f"Our model's accuracy for predicting All-Pro's: {recall_score(y_test, y_pred)*100:4.2f}%")

Our model's accuracy for predicting All-Pro's: 29.41%

Stacking Classifier

my_estimators = [
    ('DTC', DecisionTreeClassifier(max_depth = 2, random_state = 12)),
    ('RFC', best_RFC),
    ('LR', LogisticRegression(max_iter = 1000))
]

SC = StackingClassifier(estimators = my_estimators, final_estimator = best_DTC)

SC.fit(X_train, y_train)

StackingClassifier(estimators=[('DTC',
                                DecisionTreeClassifier(max_depth=2,
                                                       random_state=12)),
                               ('RFC',
                                RandomForestClassifier(max_depth=3,
                                                       n_estimators=7)),
                               ('LR', LogisticRegression(max_iter=1000))],
                   final_estimator=DecisionTreeClassifier(max_depth=1))

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

print(f"Our model's score against the testing data: {SC.score(X_test, y_test)*100:4.2f}%")

Our model's score against the testing data: 95.95%

ConfusionMatrixDisplay.from_estimator(SC, X_test, y_test)

<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7f1de53baa00>

y_pred = SC.predict(X_test)
print(f"Our model's accuracy for predicting All-Pro's: {recall_score(y_test, y_pred)*100:4.2f}%")

Our model's accuracy for predicting All-Pro's: 32.35%

df2022 = pd.read_csv('2022_nfl_qb.csv')
df2022 = df2022.loc[df2022['Pos'] == 'QB']

df2022.isna().sum()
df2022.loc[df2022['Y/C'].isna()]

	Rk	Player	Tm	Age	Pos	G	GS	QBrec	Cmp	Att	...	Rate	QBR	Sk	Yds.1	Sk%	NY/A	ANY/A	4QC	GWD	Player-additional
82	83	Chase Daniel	LAC	36	QB	2	0	NaN	0	2	...	39.6	27.4	1	1	33.3	-0.33	-0.33	NaN	NaN	DaniCh00
84	85	Chad Henne	KAN	37	QB	2	0	NaN	0	2	...	39.6	3.5	0	0	0.0	0.00	0.00	NaN	NaN	HennCh01

2 rows × 32 columns

df2022.dropna(inplace = True)

df_normalized.columns

Index(['g', 'cmp', 'cmp%', 'pass_yds', 'pass_td', 'td%', 'int%', 'pass_y/a',
       'ay/a', 'y/c', 'pass_y/g', 'rate', 'ny/a', 'any/a', 'was_ap'],
      dtype='object')

df2022.drop(columns = {'Rk', 'Player', 'Tm', 'Age', 'Pos', 'GS', 'QBrec', 'Att', 'Int', '1D', 'Lng',
                       'QBR', 'Sk', 'Yds.1', 'Sk%', '4QC', 'GWD', 'Player-additional'}, inplace = True)

df2022.columns

Index(['G', 'Cmp', 'Cmp%', 'Yds', 'TD', 'TD%', 'Int%', 'Y/A', 'AY/A', 'Y/C',
       'Y/G', 'Rate', 'NY/A', 'ANY/A'],
      dtype='object')

df2022.columns = ['g', 'cmp', 'cmp%', 'pass_yds', 'pass_td', 'td%', 'int%', 'pass_y/a', 'ay/a', 'y/c', 
                  'pass_y/g', 'rate', 'ny/a', 'any/a']

df2022['year'] = 2022
df2022_normalized = normalize_by_year(df2022)
df2022_normalized.drop(columns = {'year'}, inplace = True)

SC.predict(df2022_normalized)

array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0])

best_DTC.predict(df2022_normalized)

array([1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0])

best_RFC.predict(df2022_normalized)

array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0])

df2022_normalized

	g	cmp	cmp%	pass_yds	pass_td	td%	int%	pass_y/a	ay/a	y/c	pass_y/g	rate	ny/a	any/a
0	0.9	0.857664	0.590643	1.000000	1.000000	0.961538	0.297297	0.71875	0.772727	0.804878	1.000000	0.819797	0.767516	0.817582
1	0.9	0.843066	0.789474	0.879774	0.814815	0.826923	0.324324	0.62500	0.681818	0.536585	0.838085	0.789340	0.531847	0.613187
2	0.9	0.770073	0.497076	0.866490	0.814815	0.807692	0.513514	0.59375	0.568182	0.707317	0.819796	0.593909	0.601911	0.597802
3	0.9	0.974453	0.649123	0.844238	0.629630	0.480769	0.216216	0.21875	0.340909	0.170732	0.790210	0.500000	0.267516	0.400000
4	0.9	1.000000	0.619883	0.841913	0.481481	0.326923	0.000000	0.18750	0.340909	0.121951	0.786982	0.482234	0.248408	0.426374
5	0.9	0.773723	1.000000	0.787778	0.703704	0.807692	0.243243	0.71875	0.750000	0.487805	0.713825	0.916244	0.573248	0.665934
6	1.0	0.715328	0.374269	0.770508	0.629630	0.615385	0.459459	0.43750	0.431818	0.609756	0.583109	0.426396	0.449045	0.465934
7	0.9	0.693431	0.543860	0.738957	0.592593	0.634615	0.297297	0.53125	0.545455	0.585366	0.647660	0.586294	0.544586	0.589011
8	0.9	0.620438	0.730994	0.711724	0.629630	0.807692	0.054054	0.75000	0.863636	0.731707	0.611081	0.906091	0.646497	0.775824
9	0.9	0.791971	0.526316	0.709399	0.555556	0.500000	0.378378	0.21875	0.272727	0.243902	0.607854	0.393401	0.168790	0.276923
10	0.8	0.791971	0.713450	0.690136	0.370370	0.307692	0.648649	0.28125	0.181818	0.195122	0.699301	0.289340	0.156051	0.131868
11	1.0	0.726277	0.508772	0.686483	0.703704	0.730769	0.405405	0.31250	0.386364	0.341463	0.478214	0.502538	0.273885	0.386813
12	0.7	0.521898	0.730994	0.684822	0.666667	1.000000	0.270270	1.00000	1.000000	1.000000	0.831092	1.000000	1.000000	1.000000
13	0.9	0.729927	0.584795	0.676519	0.518519	0.519231	0.216216	0.31250	0.409091	0.317073	0.563744	0.512690	0.315287	0.432967
14	0.9	0.594891	0.497076	0.601461	0.333333	0.365385	0.270270	0.40625	0.409091	0.487805	0.462076	0.418782	0.353503	0.408791
15	0.8	0.518248	0.263158	0.584856	0.185185	0.153846	0.216216	0.40625	0.409091	0.682927	0.543841	0.276650	0.242038	0.312088
16	0.8	0.489051	0.678363	0.544670	0.481481	0.730769	0.189189	0.65625	0.727273	0.658537	0.484669	0.771574	0.649682	0.723077
17	0.9	0.569343	0.608187	0.520757	0.296296	0.346154	0.162162	0.31250	0.386364	0.292683	0.353416	0.472081	0.152866	0.307692
18	0.7	0.675182	0.625731	0.518765	0.407407	0.423077	0.324324	0.09375	0.181818	0.000000	0.562130	0.368020	0.031847	0.184615
19	0.7	0.463504	0.660819	0.483228	0.444444	0.692308	0.459459	0.53125	0.522727	0.536585	0.504572	0.598985	0.519108	0.527473
20	0.9	0.474453	0.391813	0.479907	0.518519	0.730769	0.405405	0.34375	0.409091	0.463415	0.298010	0.469543	0.292994	0.397802
21	1.0	0.405109	0.333333	0.472268	0.444444	0.692308	0.648649	0.50000	0.409091	0.731707	0.211404	0.395939	0.331210	0.338462
22	0.7	0.474453	0.368421	0.447360	0.296296	0.384615	0.756757	0.21875	0.113636	0.365854	0.446477	0.139594	0.085987	0.070330
25	0.6	0.405109	0.730994	0.387247	0.148148	0.230769	0.540541	0.46875	0.318182	0.390244	0.466380	0.365482	0.305732	0.261538
26	0.9	0.302920	0.356725	0.364995	0.370370	0.711538	0.918919	0.53125	0.340909	0.756098	0.143088	0.324873	0.168790	0.153846
27	0.6	0.430657	0.561404	0.332116	0.037037	0.000000	0.567568	0.09375	0.000000	0.073171	0.367402	0.063452	0.022293	0.000000
30	0.3	0.259124	0.380117	0.229824	0.259259	0.557692	0.540541	0.18750	0.204545	0.292683	0.628295	0.291878	0.050955	0.156044
31	0.4	0.211679	0.362573	0.214879	0.222222	0.538462	0.486486	0.31250	0.318182	0.463415	0.402905	0.347716	0.257962	0.314286
32	0.4	0.116788	0.000000	0.160080	0.037037	0.134615	0.540541	0.31250	0.181818	0.756098	0.275955	0.000000	0.159236	0.136264
33	0.5	0.076642	0.163743	0.084357	0.074074	0.326923	0.351351	0.21875	0.250000	0.512195	0.000000	0.200508	0.264331	0.318681
34	0.0	0.065693	0.181287	0.034540	0.074074	0.346154	0.351351	0.00000	0.090909	0.195122	0.908553	0.134518	0.000000	0.131868
35	0.0	0.000000	0.461988	0.020259	0.037037	0.403846	1.000000	0.53125	0.227273	0.658537	0.831630	0.175127	0.347134	0.167033
36	0.2	0.018248	0.309942	0.000000	0.000000	0.173077	0.270270	0.12500	0.181818	0.268293	0.150619	0.190355	0.095541	0.197802

best_DTC.predict(df2022_normalized[0:1])

array([1])